Syntrillium Backup Forums :: View topic

invader9000 · Posted - Mon Jul 02, 2001 9:38 am

Hi all! This may sound weird but... anyway
Imagine we have an MP3 coding software. Lat's say Fraunhofer's. Is there a way for someone to attach a "procedure" which will make the software create a log file with EVERY change that is made to the file that is encoded? I mean that if we could record the changes made on a wav that is being compressed, do you think that then the compression could be reversed in a way that "we can put the missing bits back" and re-build the original WAV file?

NO i have not graduated from the Psychological Clinic (or what ever is called). Just curious!

SteveG · Location: United Kingdom

I did an experiment once to see what the audible difference between an MP3 file and its original WAV was, by the simple expedient of subtracting one from the other. Now, I haven't tried the other half of the experiment, but on the face of it, you'd think that you could just add them back together and recreate your WAV.

Well, that's it in theory. But I suspect that in reality, it won't work very well, because it's in the nature of the multi-band masking process to introduce minute temporal errors, mainly to do with the band-splitting and analysis process. This is rather confirmed by the results of the subtractive process, which sounded distinctly 'phasey'.

If you could really break into the Fraunhofer process, you'd have to save a lot of files - one for each filtering process, and there are a lot of those. And then you'd have to reinstate each one in the same manner in which it was extracted, which would also mean breaking into the decode algorithm and changing all the data there as well. At the end of all this, would you have your WAV back? Well, possibly, but I think I'd rather have the original one, if you don't mind.

Even if the process worked reliably, the combined amount of data stored to enable the process to happen would be greater than the size of the original WAV file anyway, so what have you actually achieved?

Not a lot, I suspect! But it was an interesting way of whiling away 10 minutes while supper cooks...

Steve

invader9000 · Posted - Mon Jul 02, 2001 11:10 am

Thanks Steve for this reply. I would of course prefer the WAV to the MP3, but for storage purposes and the internet...hmmm i do not think so.

I know that we cannot re-build the sound, but just put the bits back. I was thinking of a 'bit mapper' which would track the bit positions of the uncompressed wave and another one with the compressed wave's bit map (something that will show the position of each bit). Then, an algorithm would create new bits and put them in the place where the original were, according to the WAV bitmap. That way we would not log the sound loss and recreation, but the binary data itself, which would the algorithm create, guided by a log file in ASCII probably format.

Any one seeing any light to that?

Edited by - invader9000 on 07/02/2001 11:11:56 AM

beetle · Location: USA

It won't work Invader. Mp3 encoding throws away data that cannot be recovered. That's why you cannot restore the original sound quality from an mp3. Period.

SteveG · Location: United Kingdom